- Email: [email protected]
Nitrogen-induced local spin polarization in graphene on cobalt
Journal of Magnetism and Magnetic Materials 342 (2013) 144–148
Contents lists available at SciVerse ScienceDirect
Journal of Magnetism and Magnetic Materials journal homepage: www.elsevier.com/locate/jmmm
Nitrogen-induced local spin polarization in graphene on cobalt Zhongping Chen a,b, Ling Miao a,n, Xiangshui Miao a,c,nn a b c
School of Optical and Electronic Information, Huazhong University of Science and Technology, 430074 Wuhan, China Department of Physics, The University of Texas at Austin, Austin, TX 78712, USA Wuhan National Laboratory for Optoelectronics, 430074 Wuhan, China
art ic l e i nf o
a b s t r a c t
Article history: Received 22 September 2012 Received in revised form 7 February 2013 Available online 24 April 2013
Using first principles calculations, we demonstrate an effective method to tailor the local spin configuration of graphene on Co(0 0 0 1) surface through nitrogen doping. Two different site occupancies of the N impurities are discussed with the focus on structural, electronic and magnetic properties. N induces opposite local spin polarization at the two sites through π−d Zener exchange-type hybridization with Co substrate. In addition, the induced spin polarization is energy dependent and controllable by electric field. Consequently, this structure can be applied as a spin injection source in graphene based spintronics. & 2013 Elsevier B.V. All rights reserved.
Keywords: Spintronics Graphene Doping First principle
1. Introduction Since its first experimental discovery [1], graphene has become of considerable interest to scientists in many areas due to its intriguing physical properties [2,3]. Bing comprised of light element C with weak spin–orbit coupling, graphene enjoys a large spin relaxation length (about 2 μm) at room temperature [4], which makes it an exceptional spin transport medium and advantageous over conventional semiconductors in spin electronics (spintronics) application [5]. One crucial issue for the achievement of grapheme-based spintronics is how spin-polarized electrons can be effectively injected into nonmagnetic grapheme [6,7]. Ferromagnetic (FM) contact is among the most popular spin injection methods applied in graphene spin valve devices [8–10]. Combining graphene and FM metals [11], such as Co and Ni, has been studied with the focus on interfacial structural and electronic properties [12–20], and predicted to have promising applications [21]. Nevertheless, experimental results have shown that magnetoresistance (MR) ratios in such spin valves are fairly low [8,22]. Namely, the spin injection efficiency at the graphene/FM-metal interfaces is rather limited. Therefore, it is of particular significance to find other efficient ways to inject spin into graphene. In this report, we theoretically propose one highly efficient spin injection method by implanting N impurity into a graphene sheet that is grown on Co(0 0 0 1) surface. Recently, the individual
n
Corresponding author. Tel.: +86 27 87544472. Corresponding author. E-mail addresses: [email protected] (L. Miao), [email protected]. cn (X. Miao). nn
0304-8853/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jmmm.2013.04.005
nitrogen dopants in monolayer graphene grown on a copper substrate was characterized, which strongly modified the electronic structure of nitrogen-doped grapheme [23]. So far, there have been several experimental approaches for doping nitrogen in graphene, including chemical vapor deposition [24,25], ion implantation [26] and plasma processing [27]. Simultaneously, nitrogen impurities in graphene have been studied theoretically as well [28–31]. These results have shown that nitrogen in graphene layers are almost nonmagnetic namely not spin-polarized [28,30], except some edge doping situations [30,31]. Here, by employing Co substrate, we are able to turn N impurity to highly spin-polarized states and thus create highly efficient spin injection sources in graphene. Furthermore, the spin polarization induced by N is localized, so that it can contribute to point spin injection source in nanospintronics and spin quantum bit in quantum computation. Hereby, our method is a prototype and will inspire new approaches for tailoring the local spin configuration of graphene through point defects with the help of magnetic substrates.
2. Computational details Our theoretical study is based on first principles calculations at the level of spin-polarized density functional theory (DFT) [32,33], using projector augmented wave (PAW) formalism [34] as implemented in VASP code [35]. General gradient approximation (GGA) with Perdew–Burke–Ernzerhof (PBE) exchange-correlation function [36] is employed together with a plane wave cutoff of 400 eV. Brillouin zone integration is performed on a Γ centered 6 6 1 grid using Monkhorst–Pack scheme [37]. The structural models are
Z. Chen et al. / Journal of Magnetism and Magnetic Materials 342 (2013) 144–148
constructed by substituting one C atom of a 3 3 graphene supercell with one N atom and covering the graphene sheet over a Co(0 0 0 1) substrate of five Co layers. The in-plane lattice constant a is fixed to 7.518 Å, the same as Co(0 0 0 1) lattice, and the vacuum slabs are set to more than 10 Å above the graphene sheets. Atoms in the graphene sheet as well as the upper most two Co layers are free to move during atomic relaxation, which is conducted by conjugate gradient method [38], and the structures are optimized when the force on each atom is less than 0.01 eV/Å.
3. Results and discussion First the characteristics of pure graphene adsorbed on a Co surface are briefly reviewed.1 The lattice mismatch between graphene and Co(0 0 0 1) plane is so small (1.8%) [39], that the graphene sheet is in registry with the Co surface and suffers only a very little tensile. Above the Co(0 0 0 1) surface, there are three sites marked A, B and C in Fig. 1(a). Two of them may be occupied by Graphene C atoms, resulting in three on-top registry configurations named GrAB, GrAC and GrBC. Our results show that GrAC is the most stable one with graphene–Co separation d0 ¼ 2.12 Å. A possible reason for the low spin injection efficiency at the interfaces in the spin valves, could be understood from the projected density of states (PDOS) of C atoms in GrAC , as presented in Fig. 1(b). All s, px and py components are symmetric and only pz component is asymmetric with respect to the two spins. This indicates π−d hybridization between out-of-plane π (pz ) states of graphene and d band of Co substrate. Because of their different site occupancies, the two C atoms present different pz PDOS features. CA , sitting above a surface Co, presents positive spin peaks within [−2.8, −2.2] eV, [0.1, 1.2] eV intervals, and negative spin peaks within [−1.7, −1.0] eV, [1.4, 2.2] eV intervals. In comparison, CC presents a strong positive peak within [−1.0, −0.2] eV, a strong negative peak and a week positive peak within [0.6, 1.4] eV. It is noteworthy that PDOS of both CA and CC diminish at the Fermi level (EF). Therefore the graphene is not spin-polarized [40] at EF, and the resistance across the interface is large. Next, we study the effects of N impurity in GrAC . By substituting one CA or CC with N dopant, we create two situations GrAC NA and GrAC NC as shown in Fig. 2(a) and (b), respectively. Superscript of the notation in each pane denotes the site occupancy. Subscripts (1, 2 and 3) of the C notations denote the first, second and third nearest neighbors. We calculated the binding energy ΔE ¼Etot−EGrN−ECo. Here, Etot, EGrN and ECo are energies of total structure, graphene with N impurity and Co substrate, respectively. ΔE of GrAC, GrACNA and GrACNC are −0.93 eV, −0.89 eV and −0.97 eV. NA impurity decreases the stability of graphene adsorption while NC impurity improves the stability. Fourthmore, GrACNC is more stable than GrACNA which has a relative high total energy about 0.075 eV per unit cell with one N atom. It indicates that experimentally the N impurity prefers to form at C site instead of at A site in the growth of graphene on the Co substrate with the CVD method. Some other data determining their properties are presented in Table 1. In either case, N is located above the graphene layer with the distance d1 larger than d0. The PDOS of these N impurities is spin asymmetry, especially arround EF. For GrAC NA presented in Fig. 2(c), the out-of-plane pz component becomes spin asymmetric above −5 eV. For instance, there is only a positive PDOS peak at EF, while in the negative spin channel the PDOS is almost zero at EF. A sharp positive peak and a sharp negative peak are also presented at −2.8 eV and −1.8 eV, 1
See supplemental material for more details about pure graphene on Co.
145
respectively. Similarly, NC displays spin asymmetric pz PDOS as well [see Fig. 2(d)], but the characteristics are completely different from NA . In the vicinity of EF, the NA positive peak vanishes in NC PDOS. Instead, a broad negative peak appears within [−0.6, 0.1] eV. A strong positive peak is presented within [−1.6, −0.8] eV and two other positive peaks are located around 0.3 eV and 1.6 eV. The π−d Zener exchange mechanism [41] can be employed to explain the spin asymmetry of pz states of these N impurities. In diluted magnetic semiconductors, the p band of semiconductor is mixed with the d states of the transition metal impurity ions and thereby broadened and shifted with different variations in the two spin channels around EF. In the cases discussed here, Co substrates play the roles of transition metal impurity and N act as the semiconductor. Specifically, in GrAC NA , strong hybridization is apparent as exhibited in Fig. 2(c): the positive peak at −2.8 eV and the negative peak at 1.2 eV are present both in the pz PDOS of on-top NA and the dxz , dyz and dz2 PDOS of surface CoA . While in GrAC NC , the NC negative peak around EF is the result of strong π d hybridization with dxz and dyz states of the nearest Co (CoA ) [See Fig. 2(d)]. The positive peak, as well as the negative states, within [−1.6, −0.8] eV also hybridize with Codxz and dyz states in the same energy interval. Below −5.5 eV in Fig. 2(c) and (d), where Co PDOS drops to zero, both NA and NC display spin symmetric pz PDOS because of the absence of π−d mixing. It is interesting that NA and NC have opposite local spin polarizations at EF. Actually, the NA and NC dopants on Co substrate have different adjacent atoms. NA on the top of CoA has a strong direct interaction with CoA atom of 2.25 Å NA−CoA bond through the π−d Zener exchange mechanism. Meanwhile, NC interacts directly only with three bonded C atoms, and has indirect π−d interaction with surface Co atoms. Fourthmore, it was found that NA will gain more electrons about 0.44 e than NC from the bader population analysis. So, the difference of PDOS structure and local charge doping induced by different interaction with adjacent atoms, will lead to the opposite local spin polarizations of N impurity. It may be assumed that the N impurity will introduce some changes to its neighboring graphene C and substrate Co. In fact, this is true only when N takes the place of on-site CA in GrAC NA . NA hybridizes substantially with its neighboring C in the conjugate loops and causes considerable changes to their PDOS [see Fig. 2 (c)]. Most apparently, all CC1 , CA2 and CC3 present the strong positive peaks at EF as NA does, which is not found in undoped GrAC . Also the positive peak around −2.8 eV and the negative peak around −1.8 eV exist in NA PDOS due to the hybridization between NA and CA2 that are both on-top of CoA . NA brings some changes to the CoA beneath it too. Especially for the dz2 PDOS, in Fig. 2(c) the positive peak within [−1.5, −0.4] eV and the negative peak within [0.2, 1.5] eV are both broadened and lifted compared to the CoA in GrAC NC as shown in Fig. 2(d). Nevertheless, in the more stable GrAC NC , the PDOS characteristics of neighboring C and Co are hardly altered. For example, the first (third) nearest neighbor CA1 (CA3 ) of NC is characterized by two positive PDOS peaks within [−2.8, −2.2] eV, [0.1, 1.2] eV and two negative peaks within [−1.8, −1.0] eV, [1.3, 2.2] eV [see Fig. 2(d)], which is the hallmark of CA in pure GrAC on Co. The second nearest neighbor CC2 presents similar PDOS as the CC shown in Fig. 1 too. The π−d hybridization features of undoped GrAC with Co substrate (not shown here) are also retained, such as the positive (negative) hybridization between CoA and CA1 within [−2.8, −2.2] eV ([−1.8, −1.0] eV), and the positive hybridization between CoA and CC2 around −0.5 eV [see Fig. 2(d)]. Under the influence of Co substrate, N impurity introduces high spin polarization [40] around EF to the graphene layer and thus substantially increases the initially diminishing density of states around EF in one spin channel. [Note that the vertical scales of N PDOS are twice the scale of C in Fig. 2(c) and (d)] So that the
146
Z. Chen et al. / Journal of Magnetism and Magnetic Materials 342 (2013) 144–148
Fig. 1. Pure graphene on Co. (a) Top view of 3 3 Co surface supercell with three sites marked A–C. Dark blue spheres represent surface Co atoms (CoA ), and light blue ones represent the Co atoms in the second monolayer (CoB ). (b) Spin-resolved projected density of states (PDOS) of the C atoms in GrAC configuration. The superscript of the notation in each pane denotes the site occupancy of the C atom. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 2. In-plane structural model of (a) GrAC NA and (b) GrAC NC . The gray grids represent the graphene lattice and the orange spheres represent the N impurity. (c) and (d) show spin-resolved PDOS of the N impurity and the neighboring C and Co in GrAC NA and GrAC NC , respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Z. Chen et al. / Journal of Magnetism and Magnetic Materials 342 (2013) 144–148
electrons carrying one certain spin can easily travel into the graphene but those carrying the other spin are still blocked at the interface. This mechanism exactly exhibits the function of an effective spin injection source. The spin polarization induced by N impurity is localized since the PDOS of those C far away from N are found to be unaffected. Hence, this spin injection source can be viewed as a dimensionless point compared to the graphene sheet, which possibly contributes to graphene based nanospintronics. Relatively speaking, the spin-polarized area in GrAC NA is larger than in GrAC NC , since not only NA but also all the neighboring C atoms in the three conjugate loops are positively spin-polarized at EF, and accordingly the spin-polarized current in GrAC NA is expected to be higher. As we know, Co substrate has strong contribution of the spin polarization in these models. However, the spin polarization at EF of GrAC NA is inverted with respect to the Co substrate, while GrAC NC has the same spin polarization at EF as Co [see Fig. 2(c), (d)]. Therefore which configuration is more favorable for a spin injection source still needs to be determined experimentally. Although GrAC NC is more stable according to Table 1, the formation energy difference here (0.075 eV) is rather tiny when compared to those appearing in the literature ( 1 eV). [20,23] Also, scanning tunneling microscopy (STM) study of graphene on Co has shown that the A and C sites are distinguishable [9]. Therefore it is possible to artificially control the implantation of N impurity into either site by STM, and to create the desirable spin injection source. It is possible to control the polarization direction of the spin injection source using external electric field, magnetic field or Table 1 Comparison of formation energy relative to GrAC NC [ΔEf (eV)], height of N above Co surface d1 (Å), the nearest N–Co distance d2 (Å), charge population (−e) and magnetic moment (μB ) of N (eN , μN ) as well as neighboring Co (eCo , μCo ) in GrAC NA and GrAC NC . ΔEf
d1
d2
eN
μN
eCo
μCo
GrAC NA
0.075
2.25
2.39
−2.71
0.04
0.11
1.75
GrAC NC
0
2.42
2.85
−2.27
−0.08
0.13
1.45
GrAC
0.038
2.12
2.12
–
–
0.11
1.52
147
chemical doping. It is intriguing that the spin polarization of N impurity is energy-dependent, and the fermi energy EF could be shifted with the electron/hole doping. The sharp positive peak and negative peak of NC PDOS fermi energy EF are at about 0.4 eV and −0.4 eV, respectively, as shown in revised Fig. 2(d). It means that a shift of 0.4 0.8 eV around EF is needed for achieving an obvious turn of spin polarization. Within a certain energy interval where a positive or a negative peak presents, the N states are accordingly positively or negatively spin-polarized [see Fig. 2(c) and (d)]. The spin polarization in each of these energy intervals is quite high and sometimes approaches 100%. When an electric field is applied perpendicular to the surface by a gate voltage, EF is shifted proportionally to the voltage. Therefore, the spin polarization can be tuned simply by shifting EF into a certain energy interval of the desired spin polarization. To further discuss the effects of different N site occupancies, we present in Fig. 3 the spin density and charge density in certain planes of the two GrAC NA and GrAC NC supercells. NA presents a diminishing positive spin density in Fig. 3(c), so its magnetic moment is very small (0.04 μB ). The three nearest CC keep the positive spin density and thus yield a positive net magnetic moment in the vicinity of NA . While, NC presents a considerable negative spin density in Fig. 3(d) with a magnetic moment of −0.08 μB (see Table 1) and gives rise to a negative net magnetic moment together with its three CA neighbors. Accordingly the bonding between NA and CoA is more ionic than C–Co bonding, as seen from Fig. 3(e) that the overlap of charge density between NA and CoA is slighter than that between CA and the underlying CoA . The interaction between NA and CoA is a little repulsive so that NA is located slightly above the graphene layer and CoA is slightly pushed downwards [see Fig. 3(e)], which causes the N–Co distance d2 to be a little larger than d1 in Table 1. In comparison, NC is located above the hollow site surrounded by three CoA with a larger N–Co distance (d2 ¼ 2.85 Å) in GrAC NC . So the interaction between NC and CoA is weaker and indirect as shown in the charge density in Fig. 3(f) that the overlap is very slight and, to some degree, in the in-plane direction. Since the electronegativity of N (3.04) is larger than C (2.55), NA attracts charges from adjacent C atoms, which improves the charge transfer from CoA to graphene with increasing formation energy as presented in Table 1. These findings are consistent with the PDOS properties discussed above. For instance, NA hybridizes
Fig. 3. In-plane structural model of (a) GrAC NA and (b) GrAC NC . The gray grids represent the graphene lattice and the orange spheres represent the N impurity. (c) and (d) show the spin density distribution in the planes passing through the graphene sheets in the corresponding models at left side. The red contours represent positive spin density and the green and blue ones represent higher and lower negative spin density. Contour plots of the charge density in the ð2 1 1 0Þ slices of GrAC NA and GrAC NC supercells are shown in (e) and (f), respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
148
Z. Chen et al. / Journal of Magnetism and Magnetic Materials 342 (2013) 144–148
strongly with all out of plane d states, especially dz2 states, of CoA . While NC only hybridizes with CoA dxz and dyz states, leaving dz2 states unchanged. We have also examined N impurity in GrAB and GrBC . In GrAB , NA and NB behave just similarly as NA and NC in GrAC , respectively. While in GrBC , both NB and NC show the identical symmetric PDOS regarding the two spins, and their magnetic moments, as well as moments of all the C atoms, are zero. This is because the hybridization effect disappeared with a large graphene–Co separation (d0 ¼4.18 Å). Therefore, the influence from the Co substrate is indispensable in generating the spin polarization in N-doped graphene. 4. Conclusion In summary, we have theoretically demonstrated that the local spin configuration of graphene sheet on Co(0 0 0 1) surface can be tailored by N substitution. N is positively spin-polarized at A site while negatively spin-polarized at C site. The spin polarization in either case is as high as almost 100% and largely energy dependent. Therefore they act as spin injection sources that can be controlled electrically as well as magnetically. A site substitution brings more modification to the PDOS of local atoms than C site substitution and contributes to the positive spin polarization of the surrounding C. N impurity introduces positive and negative net magnetic moments to the occupied A site and C site, respectively, which can be regarded as spin quantum bits as well. Acknowledgements This work is funded by National Natural Science Foundation of China (No. 50871043 and No. 61172003), Natural Science Foundation of Hubei Province of China (No. ZRY0871) and Innovative Foundation of Huazhong University of Science and Technology (No. 2012QN151). The computational resources are provided by Wuhan National Laboratory for Optoelectronics (WNLO). Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.jmmm.2013.04. 005.
References [1] K.S. Novoselov, A.K. Geim, S.V. Morozov, D. Jiang, Y. Zhang, S.V. Dubonos, I.V. Grigorieva, A.A. Firsov, Science 306 (2004) 666. [2] K.S. Novoselov, A.K. Geim, S.V. Morozov, D. Jiang, M.I. Katsnelson, I.V. Grigorieva, S.V. Dubonos, A.A. Firsov, Nature 438 (2005) 197. [3] Y. Zhang, Y.-W. Tan, H.L. Stormer, P. Kim, Nature 438 (2005) 201.
[4] N. Tombros, C. Jozsa, M. Popinciuc, H.T. Jonkman, B.J. van Wees, Nature 448 (2007) 571. [5] L.E. Hueso, et al., Nature 445 (2007) 410. [6] K. Tada, J. Haruyama, H.X. Yang, M. Chshiev, T. Matsui, H. Fukuyama, Physical Review Letters 107 (2011) 217203. [7] H.X. Yang, M. Chshiev, D.W. Boukhvalov, X. Waintal, S. Roche, Physical Review B: Condensed Matter 84 (2011) 214404. [8] E.W. Hill, A.K. Geim, K. Novoselov, F. Schedin, P. Blake, IEEE Transactions on Magnetics 42 (2006) 2694. [9] W. Han, W.H. Wang, K. Pi, K.M. McCreary, W. Bao, Y. Li, F. Miao, C.N. Lau, R.K. Kawakami, Physical Review Letters 102 (2009) 137205. [10] M. Ohishi, M. Shiraishi, R. Nouchi, T. Nozaki, T. Shinjo, Y. Suzuki, Japanese Journal of Applied Physics 46 (2007) L605. [11] C. Vo-Van, Z. Kassir-Bodon, H.X. Yang, J. Coraux, J. Vogel, S. Pizzini, P. BayleGuillemaud, M. Chshiev, L. Ranno, V. Guisset, P. David, V. Salvador, O. Fruchart, New Journal of Physics 12 (2010) 103040. [12] D. Eom, et al., Nano Letters 9 (2009) 2844. [13] A. Varykhalov, O. Rader, Physical Review B: Condensed Matter 80 (2009) 035437. [14] J. Maassen, W. Ji, H. Guo, Nano Letters 11 (2010) 151. [15] A. Varykhalov, J. Sánchez-Barriga, A.M. Shikin, C. Biswas, E. Vescovo, A. Rybkin, D. Marchenko, O. Rader, Physical Review Letters 101 (2008) 157601. [16] G. Giovannetti, P.A. Khomyakov, G. Brocks, V.M. Karpan, J. van den Brink, P.J. Kelly, Physical Review Letters 101 (2008) 026803. [17] P.A. Khomyakov, G. Giovannetti, P.C. Rusu, G. Brocks, J. van den Brink, P.J. Kelly, Physical Review B: Condensed Matter 79 (2009) 195425. [18] Y.L. Mao, J. Yuan, J.X. Zhong, Journal of Physics: Condensed Matter 20 (2008) 115209. [19] Y.L. Mao, J.X. Zhong, Nanotechnology 19 (2008) 205708. [20] Z.L. Zhang, Y.P. Chen, Y.E. Xie, M. Zhang, J.X. Zhong, Journal of Physics D: Applied Physics 44 (2011) 215403. [21] V.W. Brar, R. Decker, H.M. Solowan, et al., Nature Physics 7 (2011) 43. [22] L. Brey, H.A. Fertig, Physical Review B: Condensed Matter 76 (2007) 205435. [23] L.Y. Zhao, R. He, K.T. Rim, et al., Science 333 (2011) 999. [24] D. Wei, Y. Liu, Y. Wang, H. Zhang, L. Huang, G. Yu, Nano Letters 9 (2009) 1752. [25] A.L.M. Reddy, A. Srivastava, S.R. Gowda, H. Gullapalli, M. Dubey, P.M. Ajayan, ACS Nano 4 (2010) 6337. [26] U. Bangert, A. Bleloch, M.H. Gass, A. Seepujak, J. van den Berg, Physical Review B: Condensed Matter 81 (2010) 245423. [27] H.M. Jeong, J.W. Lee, W.H. Shin, Y.J. Choi, H.J. Shin, J.K. Kang, J.W. Choi, Nano Letters 11 (2011) 2472. [28] S.-F. Huang, K. Terakura, T. Ozaki, T. Ikeda, M. Boero, M. Oshima, J.-i. Ozaki, S. Miyata, Physical Review B: Condensed Matter 80 (2009) 235410. [29] M.G. Menezes, R.B. Capaz, J.L.B. Faria, Physical Review B: Condensed Matter 82 (2010) 245414. [30] E. Cruz-Silva, Z.M. Barnett, B.G. Sumpter, V. Meunier, Physical Review B: Condensed Matter 83 (2011) 155445. [31] Y. Li, Z. Zhou, P. Shen, Z. Chen, ACS Nano 3 (2009) 1952. [32] P. Hohenberg, W. Kohn, Physical Review B: Condensed Matter 136 (1964) 864. [33] W. Kohn, L.J. Sham, Physical Review A: Atomic, Molecular, and Optical Physics 140 (1965) 1133. [34] G. Kresse, D. Joubert, Physical Review B: Condensed Matter 59 (1999) 1758. [35] G. Kresse, J. Furthmüller, Physical Review B: Condensed Matter 54 (1996) 11169. [36] J.P. Perdew, K. Burke, M. Ernzerhof, Physical Review Letters 77 (1996) 3865. [37] H.J. Monkhorst, J.D. Pack, Physical Review B: Condensed Matter 13 (1976) 5188. [38] M.C. Payne, M.P. Teter, D.C. Allan, T.A. Arias, J.D. Joannopoulos, Reviews of Modern Physics 64 (1992) 1045. [39] V.M. Karpan, G. Giovannetti, P.A. Khomyakov, M. Talanana, A.A. Starikov, M. Zwierzycki, J. van den Brink, G. Brocks, P.J. Kelly, Physical Review Letters 99 (2007) 176602. [40] The spin polarization is defined as (D↑–D↓)/(D↑–D↓), where D↑ and D↓ are integration of positive and negative PDOS around EF or over a certain energy interval if specified. [41] J. Kanamori, K. Terakura, Journal of the Physical Society of Japan 70 (2001) 1433.
Contents lists available at SciVerse ScienceDirect
Journal of Magnetism and Magnetic Materials journal homepage: www.elsevier.com/locate/jmmm
Nitrogen-induced local spin polarization in graphene on cobalt Zhongping Chen a,b, Ling Miao a,n, Xiangshui Miao a,c,nn a b c
School of Optical and Electronic Information, Huazhong University of Science and Technology, 430074 Wuhan, China Department of Physics, The University of Texas at Austin, Austin, TX 78712, USA Wuhan National Laboratory for Optoelectronics, 430074 Wuhan, China
art ic l e i nf o
a b s t r a c t
Article history: Received 22 September 2012 Received in revised form 7 February 2013 Available online 24 April 2013
Using first principles calculations, we demonstrate an effective method to tailor the local spin configuration of graphene on Co(0 0 0 1) surface through nitrogen doping. Two different site occupancies of the N impurities are discussed with the focus on structural, electronic and magnetic properties. N induces opposite local spin polarization at the two sites through π−d Zener exchange-type hybridization with Co substrate. In addition, the induced spin polarization is energy dependent and controllable by electric field. Consequently, this structure can be applied as a spin injection source in graphene based spintronics. & 2013 Elsevier B.V. All rights reserved.
Keywords: Spintronics Graphene Doping First principle
1. Introduction Since its first experimental discovery [1], graphene has become of considerable interest to scientists in many areas due to its intriguing physical properties [2,3]. Bing comprised of light element C with weak spin–orbit coupling, graphene enjoys a large spin relaxation length (about 2 μm) at room temperature [4], which makes it an exceptional spin transport medium and advantageous over conventional semiconductors in spin electronics (spintronics) application [5]. One crucial issue for the achievement of grapheme-based spintronics is how spin-polarized electrons can be effectively injected into nonmagnetic grapheme [6,7]. Ferromagnetic (FM) contact is among the most popular spin injection methods applied in graphene spin valve devices [8–10]. Combining graphene and FM metals [11], such as Co and Ni, has been studied with the focus on interfacial structural and electronic properties [12–20], and predicted to have promising applications [21]. Nevertheless, experimental results have shown that magnetoresistance (MR) ratios in such spin valves are fairly low [8,22]. Namely, the spin injection efficiency at the graphene/FM-metal interfaces is rather limited. Therefore, it is of particular significance to find other efficient ways to inject spin into graphene. In this report, we theoretically propose one highly efficient spin injection method by implanting N impurity into a graphene sheet that is grown on Co(0 0 0 1) surface. Recently, the individual
n
Corresponding author. Tel.: +86 27 87544472. Corresponding author. E-mail addresses: [email protected] (L. Miao), [email protected]. cn (X. Miao). nn
0304-8853/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jmmm.2013.04.005
nitrogen dopants in monolayer graphene grown on a copper substrate was characterized, which strongly modified the electronic structure of nitrogen-doped grapheme [23]. So far, there have been several experimental approaches for doping nitrogen in graphene, including chemical vapor deposition [24,25], ion implantation [26] and plasma processing [27]. Simultaneously, nitrogen impurities in graphene have been studied theoretically as well [28–31]. These results have shown that nitrogen in graphene layers are almost nonmagnetic namely not spin-polarized [28,30], except some edge doping situations [30,31]. Here, by employing Co substrate, we are able to turn N impurity to highly spin-polarized states and thus create highly efficient spin injection sources in graphene. Furthermore, the spin polarization induced by N is localized, so that it can contribute to point spin injection source in nanospintronics and spin quantum bit in quantum computation. Hereby, our method is a prototype and will inspire new approaches for tailoring the local spin configuration of graphene through point defects with the help of magnetic substrates.
2. Computational details Our theoretical study is based on first principles calculations at the level of spin-polarized density functional theory (DFT) [32,33], using projector augmented wave (PAW) formalism [34] as implemented in VASP code [35]. General gradient approximation (GGA) with Perdew–Burke–Ernzerhof (PBE) exchange-correlation function [36] is employed together with a plane wave cutoff of 400 eV. Brillouin zone integration is performed on a Γ centered 6 6 1 grid using Monkhorst–Pack scheme [37]. The structural models are
Z. Chen et al. / Journal of Magnetism and Magnetic Materials 342 (2013) 144–148
constructed by substituting one C atom of a 3 3 graphene supercell with one N atom and covering the graphene sheet over a Co(0 0 0 1) substrate of five Co layers. The in-plane lattice constant a is fixed to 7.518 Å, the same as Co(0 0 0 1) lattice, and the vacuum slabs are set to more than 10 Å above the graphene sheets. Atoms in the graphene sheet as well as the upper most two Co layers are free to move during atomic relaxation, which is conducted by conjugate gradient method [38], and the structures are optimized when the force on each atom is less than 0.01 eV/Å.
3. Results and discussion First the characteristics of pure graphene adsorbed on a Co surface are briefly reviewed.1 The lattice mismatch between graphene and Co(0 0 0 1) plane is so small (1.8%) [39], that the graphene sheet is in registry with the Co surface and suffers only a very little tensile. Above the Co(0 0 0 1) surface, there are three sites marked A, B and C in Fig. 1(a). Two of them may be occupied by Graphene C atoms, resulting in three on-top registry configurations named GrAB, GrAC and GrBC. Our results show that GrAC is the most stable one with graphene–Co separation d0 ¼ 2.12 Å. A possible reason for the low spin injection efficiency at the interfaces in the spin valves, could be understood from the projected density of states (PDOS) of C atoms in GrAC , as presented in Fig. 1(b). All s, px and py components are symmetric and only pz component is asymmetric with respect to the two spins. This indicates π−d hybridization between out-of-plane π (pz ) states of graphene and d band of Co substrate. Because of their different site occupancies, the two C atoms present different pz PDOS features. CA , sitting above a surface Co, presents positive spin peaks within [−2.8, −2.2] eV, [0.1, 1.2] eV intervals, and negative spin peaks within [−1.7, −1.0] eV, [1.4, 2.2] eV intervals. In comparison, CC presents a strong positive peak within [−1.0, −0.2] eV, a strong negative peak and a week positive peak within [0.6, 1.4] eV. It is noteworthy that PDOS of both CA and CC diminish at the Fermi level (EF). Therefore the graphene is not spin-polarized [40] at EF, and the resistance across the interface is large. Next, we study the effects of N impurity in GrAC . By substituting one CA or CC with N dopant, we create two situations GrAC NA and GrAC NC as shown in Fig. 2(a) and (b), respectively. Superscript of the notation in each pane denotes the site occupancy. Subscripts (1, 2 and 3) of the C notations denote the first, second and third nearest neighbors. We calculated the binding energy ΔE ¼Etot−EGrN−ECo. Here, Etot, EGrN and ECo are energies of total structure, graphene with N impurity and Co substrate, respectively. ΔE of GrAC, GrACNA and GrACNC are −0.93 eV, −0.89 eV and −0.97 eV. NA impurity decreases the stability of graphene adsorption while NC impurity improves the stability. Fourthmore, GrACNC is more stable than GrACNA which has a relative high total energy about 0.075 eV per unit cell with one N atom. It indicates that experimentally the N impurity prefers to form at C site instead of at A site in the growth of graphene on the Co substrate with the CVD method. Some other data determining their properties are presented in Table 1. In either case, N is located above the graphene layer with the distance d1 larger than d0. The PDOS of these N impurities is spin asymmetry, especially arround EF. For GrAC NA presented in Fig. 2(c), the out-of-plane pz component becomes spin asymmetric above −5 eV. For instance, there is only a positive PDOS peak at EF, while in the negative spin channel the PDOS is almost zero at EF. A sharp positive peak and a sharp negative peak are also presented at −2.8 eV and −1.8 eV, 1
See supplemental material for more details about pure graphene on Co.
145
respectively. Similarly, NC displays spin asymmetric pz PDOS as well [see Fig. 2(d)], but the characteristics are completely different from NA . In the vicinity of EF, the NA positive peak vanishes in NC PDOS. Instead, a broad negative peak appears within [−0.6, 0.1] eV. A strong positive peak is presented within [−1.6, −0.8] eV and two other positive peaks are located around 0.3 eV and 1.6 eV. The π−d Zener exchange mechanism [41] can be employed to explain the spin asymmetry of pz states of these N impurities. In diluted magnetic semiconductors, the p band of semiconductor is mixed with the d states of the transition metal impurity ions and thereby broadened and shifted with different variations in the two spin channels around EF. In the cases discussed here, Co substrates play the roles of transition metal impurity and N act as the semiconductor. Specifically, in GrAC NA , strong hybridization is apparent as exhibited in Fig. 2(c): the positive peak at −2.8 eV and the negative peak at 1.2 eV are present both in the pz PDOS of on-top NA and the dxz , dyz and dz2 PDOS of surface CoA . While in GrAC NC , the NC negative peak around EF is the result of strong π d hybridization with dxz and dyz states of the nearest Co (CoA ) [See Fig. 2(d)]. The positive peak, as well as the negative states, within [−1.6, −0.8] eV also hybridize with Codxz and dyz states in the same energy interval. Below −5.5 eV in Fig. 2(c) and (d), where Co PDOS drops to zero, both NA and NC display spin symmetric pz PDOS because of the absence of π−d mixing. It is interesting that NA and NC have opposite local spin polarizations at EF. Actually, the NA and NC dopants on Co substrate have different adjacent atoms. NA on the top of CoA has a strong direct interaction with CoA atom of 2.25 Å NA−CoA bond through the π−d Zener exchange mechanism. Meanwhile, NC interacts directly only with three bonded C atoms, and has indirect π−d interaction with surface Co atoms. Fourthmore, it was found that NA will gain more electrons about 0.44 e than NC from the bader population analysis. So, the difference of PDOS structure and local charge doping induced by different interaction with adjacent atoms, will lead to the opposite local spin polarizations of N impurity. It may be assumed that the N impurity will introduce some changes to its neighboring graphene C and substrate Co. In fact, this is true only when N takes the place of on-site CA in GrAC NA . NA hybridizes substantially with its neighboring C in the conjugate loops and causes considerable changes to their PDOS [see Fig. 2 (c)]. Most apparently, all CC1 , CA2 and CC3 present the strong positive peaks at EF as NA does, which is not found in undoped GrAC . Also the positive peak around −2.8 eV and the negative peak around −1.8 eV exist in NA PDOS due to the hybridization between NA and CA2 that are both on-top of CoA . NA brings some changes to the CoA beneath it too. Especially for the dz2 PDOS, in Fig. 2(c) the positive peak within [−1.5, −0.4] eV and the negative peak within [0.2, 1.5] eV are both broadened and lifted compared to the CoA in GrAC NC as shown in Fig. 2(d). Nevertheless, in the more stable GrAC NC , the PDOS characteristics of neighboring C and Co are hardly altered. For example, the first (third) nearest neighbor CA1 (CA3 ) of NC is characterized by two positive PDOS peaks within [−2.8, −2.2] eV, [0.1, 1.2] eV and two negative peaks within [−1.8, −1.0] eV, [1.3, 2.2] eV [see Fig. 2(d)], which is the hallmark of CA in pure GrAC on Co. The second nearest neighbor CC2 presents similar PDOS as the CC shown in Fig. 1 too. The π−d hybridization features of undoped GrAC with Co substrate (not shown here) are also retained, such as the positive (negative) hybridization between CoA and CA1 within [−2.8, −2.2] eV ([−1.8, −1.0] eV), and the positive hybridization between CoA and CC2 around −0.5 eV [see Fig. 2(d)]. Under the influence of Co substrate, N impurity introduces high spin polarization [40] around EF to the graphene layer and thus substantially increases the initially diminishing density of states around EF in one spin channel. [Note that the vertical scales of N PDOS are twice the scale of C in Fig. 2(c) and (d)] So that the
146
Z. Chen et al. / Journal of Magnetism and Magnetic Materials 342 (2013) 144–148
Fig. 1. Pure graphene on Co. (a) Top view of 3 3 Co surface supercell with three sites marked A–C. Dark blue spheres represent surface Co atoms (CoA ), and light blue ones represent the Co atoms in the second monolayer (CoB ). (b) Spin-resolved projected density of states (PDOS) of the C atoms in GrAC configuration. The superscript of the notation in each pane denotes the site occupancy of the C atom. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 2. In-plane structural model of (a) GrAC NA and (b) GrAC NC . The gray grids represent the graphene lattice and the orange spheres represent the N impurity. (c) and (d) show spin-resolved PDOS of the N impurity and the neighboring C and Co in GrAC NA and GrAC NC , respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Z. Chen et al. / Journal of Magnetism and Magnetic Materials 342 (2013) 144–148
electrons carrying one certain spin can easily travel into the graphene but those carrying the other spin are still blocked at the interface. This mechanism exactly exhibits the function of an effective spin injection source. The spin polarization induced by N impurity is localized since the PDOS of those C far away from N are found to be unaffected. Hence, this spin injection source can be viewed as a dimensionless point compared to the graphene sheet, which possibly contributes to graphene based nanospintronics. Relatively speaking, the spin-polarized area in GrAC NA is larger than in GrAC NC , since not only NA but also all the neighboring C atoms in the three conjugate loops are positively spin-polarized at EF, and accordingly the spin-polarized current in GrAC NA is expected to be higher. As we know, Co substrate has strong contribution of the spin polarization in these models. However, the spin polarization at EF of GrAC NA is inverted with respect to the Co substrate, while GrAC NC has the same spin polarization at EF as Co [see Fig. 2(c), (d)]. Therefore which configuration is more favorable for a spin injection source still needs to be determined experimentally. Although GrAC NC is more stable according to Table 1, the formation energy difference here (0.075 eV) is rather tiny when compared to those appearing in the literature ( 1 eV). [20,23] Also, scanning tunneling microscopy (STM) study of graphene on Co has shown that the A and C sites are distinguishable [9]. Therefore it is possible to artificially control the implantation of N impurity into either site by STM, and to create the desirable spin injection source. It is possible to control the polarization direction of the spin injection source using external electric field, magnetic field or Table 1 Comparison of formation energy relative to GrAC NC [ΔEf (eV)], height of N above Co surface d1 (Å), the nearest N–Co distance d2 (Å), charge population (−e) and magnetic moment (μB ) of N (eN , μN ) as well as neighboring Co (eCo , μCo ) in GrAC NA and GrAC NC . ΔEf
d1
d2
eN
μN
eCo
μCo
GrAC NA
0.075
2.25
2.39
−2.71
0.04
0.11
1.75
GrAC NC
0
2.42
2.85
−2.27
−0.08
0.13
1.45
GrAC
0.038
2.12
2.12
–
–
0.11
1.52
147
chemical doping. It is intriguing that the spin polarization of N impurity is energy-dependent, and the fermi energy EF could be shifted with the electron/hole doping. The sharp positive peak and negative peak of NC PDOS fermi energy EF are at about 0.4 eV and −0.4 eV, respectively, as shown in revised Fig. 2(d). It means that a shift of 0.4 0.8 eV around EF is needed for achieving an obvious turn of spin polarization. Within a certain energy interval where a positive or a negative peak presents, the N states are accordingly positively or negatively spin-polarized [see Fig. 2(c) and (d)]. The spin polarization in each of these energy intervals is quite high and sometimes approaches 100%. When an electric field is applied perpendicular to the surface by a gate voltage, EF is shifted proportionally to the voltage. Therefore, the spin polarization can be tuned simply by shifting EF into a certain energy interval of the desired spin polarization. To further discuss the effects of different N site occupancies, we present in Fig. 3 the spin density and charge density in certain planes of the two GrAC NA and GrAC NC supercells. NA presents a diminishing positive spin density in Fig. 3(c), so its magnetic moment is very small (0.04 μB ). The three nearest CC keep the positive spin density and thus yield a positive net magnetic moment in the vicinity of NA . While, NC presents a considerable negative spin density in Fig. 3(d) with a magnetic moment of −0.08 μB (see Table 1) and gives rise to a negative net magnetic moment together with its three CA neighbors. Accordingly the bonding between NA and CoA is more ionic than C–Co bonding, as seen from Fig. 3(e) that the overlap of charge density between NA and CoA is slighter than that between CA and the underlying CoA . The interaction between NA and CoA is a little repulsive so that NA is located slightly above the graphene layer and CoA is slightly pushed downwards [see Fig. 3(e)], which causes the N–Co distance d2 to be a little larger than d1 in Table 1. In comparison, NC is located above the hollow site surrounded by three CoA with a larger N–Co distance (d2 ¼ 2.85 Å) in GrAC NC . So the interaction between NC and CoA is weaker and indirect as shown in the charge density in Fig. 3(f) that the overlap is very slight and, to some degree, in the in-plane direction. Since the electronegativity of N (3.04) is larger than C (2.55), NA attracts charges from adjacent C atoms, which improves the charge transfer from CoA to graphene with increasing formation energy as presented in Table 1. These findings are consistent with the PDOS properties discussed above. For instance, NA hybridizes
Fig. 3. In-plane structural model of (a) GrAC NA and (b) GrAC NC . The gray grids represent the graphene lattice and the orange spheres represent the N impurity. (c) and (d) show the spin density distribution in the planes passing through the graphene sheets in the corresponding models at left side. The red contours represent positive spin density and the green and blue ones represent higher and lower negative spin density. Contour plots of the charge density in the ð2 1 1 0Þ slices of GrAC NA and GrAC NC supercells are shown in (e) and (f), respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
148
Z. Chen et al. / Journal of Magnetism and Magnetic Materials 342 (2013) 144–148
strongly with all out of plane d states, especially dz2 states, of CoA . While NC only hybridizes with CoA dxz and dyz states, leaving dz2 states unchanged. We have also examined N impurity in GrAB and GrBC . In GrAB , NA and NB behave just similarly as NA and NC in GrAC , respectively. While in GrBC , both NB and NC show the identical symmetric PDOS regarding the two spins, and their magnetic moments, as well as moments of all the C atoms, are zero. This is because the hybridization effect disappeared with a large graphene–Co separation (d0 ¼4.18 Å). Therefore, the influence from the Co substrate is indispensable in generating the spin polarization in N-doped graphene. 4. Conclusion In summary, we have theoretically demonstrated that the local spin configuration of graphene sheet on Co(0 0 0 1) surface can be tailored by N substitution. N is positively spin-polarized at A site while negatively spin-polarized at C site. The spin polarization in either case is as high as almost 100% and largely energy dependent. Therefore they act as spin injection sources that can be controlled electrically as well as magnetically. A site substitution brings more modification to the PDOS of local atoms than C site substitution and contributes to the positive spin polarization of the surrounding C. N impurity introduces positive and negative net magnetic moments to the occupied A site and C site, respectively, which can be regarded as spin quantum bits as well. Acknowledgements This work is funded by National Natural Science Foundation of China (No. 50871043 and No. 61172003), Natural Science Foundation of Hubei Province of China (No. ZRY0871) and Innovative Foundation of Huazhong University of Science and Technology (No. 2012QN151). The computational resources are provided by Wuhan National Laboratory for Optoelectronics (WNLO). Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.jmmm.2013.04. 005.
References [1] K.S. Novoselov, A.K. Geim, S.V. Morozov, D. Jiang, Y. Zhang, S.V. Dubonos, I.V. Grigorieva, A.A. Firsov, Science 306 (2004) 666. [2] K.S. Novoselov, A.K. Geim, S.V. Morozov, D. Jiang, M.I. Katsnelson, I.V. Grigorieva, S.V. Dubonos, A.A. Firsov, Nature 438 (2005) 197. [3] Y. Zhang, Y.-W. Tan, H.L. Stormer, P. Kim, Nature 438 (2005) 201.
[4] N. Tombros, C. Jozsa, M. Popinciuc, H.T. Jonkman, B.J. van Wees, Nature 448 (2007) 571. [5] L.E. Hueso, et al., Nature 445 (2007) 410. [6] K. Tada, J. Haruyama, H.X. Yang, M. Chshiev, T. Matsui, H. Fukuyama, Physical Review Letters 107 (2011) 217203. [7] H.X. Yang, M. Chshiev, D.W. Boukhvalov, X. Waintal, S. Roche, Physical Review B: Condensed Matter 84 (2011) 214404. [8] E.W. Hill, A.K. Geim, K. Novoselov, F. Schedin, P. Blake, IEEE Transactions on Magnetics 42 (2006) 2694. [9] W. Han, W.H. Wang, K. Pi, K.M. McCreary, W. Bao, Y. Li, F. Miao, C.N. Lau, R.K. Kawakami, Physical Review Letters 102 (2009) 137205. [10] M. Ohishi, M. Shiraishi, R. Nouchi, T. Nozaki, T. Shinjo, Y. Suzuki, Japanese Journal of Applied Physics 46 (2007) L605. [11] C. Vo-Van, Z. Kassir-Bodon, H.X. Yang, J. Coraux, J. Vogel, S. Pizzini, P. BayleGuillemaud, M. Chshiev, L. Ranno, V. Guisset, P. David, V. Salvador, O. Fruchart, New Journal of Physics 12 (2010) 103040. [12] D. Eom, et al., Nano Letters 9 (2009) 2844. [13] A. Varykhalov, O. Rader, Physical Review B: Condensed Matter 80 (2009) 035437. [14] J. Maassen, W. Ji, H. Guo, Nano Letters 11 (2010) 151. [15] A. Varykhalov, J. Sánchez-Barriga, A.M. Shikin, C. Biswas, E. Vescovo, A. Rybkin, D. Marchenko, O. Rader, Physical Review Letters 101 (2008) 157601. [16] G. Giovannetti, P.A. Khomyakov, G. Brocks, V.M. Karpan, J. van den Brink, P.J. Kelly, Physical Review Letters 101 (2008) 026803. [17] P.A. Khomyakov, G. Giovannetti, P.C. Rusu, G. Brocks, J. van den Brink, P.J. Kelly, Physical Review B: Condensed Matter 79 (2009) 195425. [18] Y.L. Mao, J. Yuan, J.X. Zhong, Journal of Physics: Condensed Matter 20 (2008) 115209. [19] Y.L. Mao, J.X. Zhong, Nanotechnology 19 (2008) 205708. [20] Z.L. Zhang, Y.P. Chen, Y.E. Xie, M. Zhang, J.X. Zhong, Journal of Physics D: Applied Physics 44 (2011) 215403. [21] V.W. Brar, R. Decker, H.M. Solowan, et al., Nature Physics 7 (2011) 43. [22] L. Brey, H.A. Fertig, Physical Review B: Condensed Matter 76 (2007) 205435. [23] L.Y. Zhao, R. He, K.T. Rim, et al., Science 333 (2011) 999. [24] D. Wei, Y. Liu, Y. Wang, H. Zhang, L. Huang, G. Yu, Nano Letters 9 (2009) 1752. [25] A.L.M. Reddy, A. Srivastava, S.R. Gowda, H. Gullapalli, M. Dubey, P.M. Ajayan, ACS Nano 4 (2010) 6337. [26] U. Bangert, A. Bleloch, M.H. Gass, A. Seepujak, J. van den Berg, Physical Review B: Condensed Matter 81 (2010) 245423. [27] H.M. Jeong, J.W. Lee, W.H. Shin, Y.J. Choi, H.J. Shin, J.K. Kang, J.W. Choi, Nano Letters 11 (2011) 2472. [28] S.-F. Huang, K. Terakura, T. Ozaki, T. Ikeda, M. Boero, M. Oshima, J.-i. Ozaki, S. Miyata, Physical Review B: Condensed Matter 80 (2009) 235410. [29] M.G. Menezes, R.B. Capaz, J.L.B. Faria, Physical Review B: Condensed Matter 82 (2010) 245414. [30] E. Cruz-Silva, Z.M. Barnett, B.G. Sumpter, V. Meunier, Physical Review B: Condensed Matter 83 (2011) 155445. [31] Y. Li, Z. Zhou, P. Shen, Z. Chen, ACS Nano 3 (2009) 1952. [32] P. Hohenberg, W. Kohn, Physical Review B: Condensed Matter 136 (1964) 864. [33] W. Kohn, L.J. Sham, Physical Review A: Atomic, Molecular, and Optical Physics 140 (1965) 1133. [34] G. Kresse, D. Joubert, Physical Review B: Condensed Matter 59 (1999) 1758. [35] G. Kresse, J. Furthmüller, Physical Review B: Condensed Matter 54 (1996) 11169. [36] J.P. Perdew, K. Burke, M. Ernzerhof, Physical Review Letters 77 (1996) 3865. [37] H.J. Monkhorst, J.D. Pack, Physical Review B: Condensed Matter 13 (1976) 5188. [38] M.C. Payne, M.P. Teter, D.C. Allan, T.A. Arias, J.D. Joannopoulos, Reviews of Modern Physics 64 (1992) 1045. [39] V.M. Karpan, G. Giovannetti, P.A. Khomyakov, M. Talanana, A.A. Starikov, M. Zwierzycki, J. van den Brink, G. Brocks, P.J. Kelly, Physical Review Letters 99 (2007) 176602. [40] The spin polarization is defined as (D↑–D↓)/(D↑–D↓), where D↑ and D↓ are integration of positive and negative PDOS around EF or over a certain energy interval if specified. [41] J. Kanamori, K. Terakura, Journal of the Physical Society of Japan 70 (2001) 1433.